Abstract
To detect and classify objects contained in real images, acquired in unconstrained environments, is a challenging problem in computer vision, which complexity makes unfeasible the design of handcrafted solutions. In this chapter, the object detection problem is introduced, highlighting the main issues and challenges, and providing a basic introduction to the main concepts. Once the problem is formulated, a feature based approach is adopted for traffic sign detection, introducing the basic concepts of the machine learning framework and some bio-inspired features. Learning algorithms are explained in order to obtain good detectors using a rich description of traffic sign instances. Using the context of classical windowing detection strategies, this chapter introduces an evolutionary approach to feature selection which allows building detectors using feature sets with large cardinalities.
Original language | English |
---|---|
Title of host publication | SpringerBriefs in Computer Science |
Number of pages | 38 |
Publisher | Springer VS |
Publication date | 2011 |
Edition | 9781447122449 |
Pages | 15-52 |
DOIs | |
Publication status | Published - 2011 |
Externally published | Yes |
Series | SpringerBriefs in Computer Science |
---|---|
Number | 9781447122449 |
Volume | 0 |
ISSN | 2191-5768 |
Bibliographical note
Publisher Copyright:© Sergio Escalera 2011.
Keywords
- Adaboost detection
- Cascade of classifiers
- Evolutionary computation
- Haar-like features
- Integral image
- Traffic sign detection